CN112217198B - Photovoltaic power generation management method for multi-step graph neural network self-adaptive dynamic planning - Google Patents

Photovoltaic power generation management method for multi-step graph neural network self-adaptive dynamic planning Download PDF

Info

Publication number
CN112217198B
CN112217198B CN202010900514.8A CN202010900514A CN112217198B CN 112217198 B CN112217198 B CN 112217198B CN 202010900514 A CN202010900514 A CN 202010900514A CN 112217198 B CN112217198 B CN 112217198B
Authority
CN
China
Prior art keywords
power generation
photovoltaic power
graph
neural network
network module
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010900514.8A
Other languages
Chinese (zh)
Other versions
CN112217198A (en
Inventor
殷林飞
杨自豪
高放
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangxi University
Original Assignee
Guangxi University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangxi University filed Critical Guangxi University
Priority to CN202010900514.8A priority Critical patent/CN112217198B/en
Publication of CN112217198A publication Critical patent/CN112217198A/en
Application granted granted Critical
Publication of CN112217198B publication Critical patent/CN112217198B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06312—Adjustment or analysis of established resource schedule, e.g. resource or task levelling, or dynamic rescheduling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/008—Circuit arrangements for ac mains or ac distribution networks involving trading of energy or energy transmission rights
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/28—Arrangements for balancing of the load in a network by storage of energy
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
    • H02J3/381—Dispersed generators
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
    • H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
    • H02J2300/20—The dispersed energy generation being of renewable origin
    • H02J2300/22—The renewable source being solar energy
    • H02J2300/24—The renewable source being solar energy of photovoltaic origin
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
    • Y02E40/70—Smart grids as climate change mitigation technology in the energy generation sector
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E70/00—Other energy conversion or management systems reducing GHG emissions
    • Y02E70/30—Systems combining energy storage with energy generation of non-fossil origin

Abstract

The invention provides a photovoltaic power generation management method based on multi-step graph neural network self-adaptive dynamic planning. The method provided by the invention comprises three basic modules, namely a graph model network module, a graph evaluation network module and a graph control network module, wherein the core structure of each module is a graph neural network. Firstly, the multi-step graph neural network self-adaptive dynamic planning method forms a large amount of photovoltaic power generation data into a visual picture and establishes a photovoltaic power generation data model. And then, the formed photovoltaic power generation data model is used for carrying out real-time control and optimization on the photovoltaic power generation system through a multi-step self-adaptive dynamic programming method, so that the photovoltaic power generation efficiency is improved.

Description

Photovoltaic power generation management method for multi-step graph neural network self-adaptive dynamic planning
Technical Field
The invention belongs to the field of new energy of an electric power system, and relates to a photovoltaic power generation management method which is suitable for multi-target planning and control of photovoltaic power generation in household energy management.
Background
Household photovoltaic is a small photovoltaic power generation device arranged on a household roof, and the household photovoltaic power generation device is a better choice in distributed power generation because the household photovoltaic power generation device does not waste land resources, has green and environment-friendly energy sources, is low in cost and has less power transmission loss. However, when the household is connected with photovoltaic power generation, the problem of power fluctuation is caused. At the same time, the electrical energy generated by the household photovoltaic inevitably leads to electrical energy distribution problems.
At present, a plurality of scholars at home and abroad develop researches on household photovoltaic power generation and household energy management. In the aspect of household photovoltaic power generation, a student provides an approximately optimal storage control algorithm, and combines a residential photovoltaic power generation system and an energy storage system to peak load curves of a power system, so that the consumption of users is reduced. In order to solve the problems of electric energy quality and reliability caused by popularization of a roof photovoltaic system, particularly the problem that photovoltaic power generation exceeds the household demand, a method for determining active power and reactive power set points of a photovoltaic inverter in a residential system is provided by a scholart. The students adjust the household power utilization plan through the establishment of the load aggregation quotient mathematical model so as to reduce the power expenditure of the users. In the aspect of energy management, a participation form of a virtual community which aims at automation and intelligence can manage production consumers in an energy sharing process in a smart power grid, and higher economic benefit is obtained.
In summary, in the aspect of household photovoltaic power generation, the above research mostly adopts a way that the energy storage device is matched with the household photovoltaic inverter for reactive power regulation to solve the problem. But the actual cost of this approach is relatively high and the photovoltaic utilization is also relatively low. In energy management, many documents adopt centralized management or control of a plurality of households, and the effect of reducing load fluctuations in household units cannot be obtained even if the electricity consumption of a single household is not high. And if the photovoltaic power generation for the user adopts a mode of nearby consumption and other grid connection, the problem of unbalanced power can be caused.
On the basis of the research, the invention provides a photovoltaic power generation management mechanism based on an artificial intelligence method, and the method has the advantages of forming a large amount of photovoltaic power generation data into visual pictures, reducing power fluctuation and saving electric power cost, and can effectively solve the problem of multi-target planning and control of household photovoltaic in household energy management.
Disclosure of Invention
The invention provides a photovoltaic power generation management method of a multi-step graph neural network self-adaptive dynamic planning. The method applies the self-adaptive dynamic planning method to photovoltaic power generation management, and simultaneously provides a new control and optimization method, namely a multi-step self-adaptive dynamic planning method. Secondly, adding a graph neural network on the basis of the multi-step adaptive dynamic planning to form a multi-step graph neural network adaptive dynamic planning method.
When a user accesses a household by photovoltaic power generation, power fluctuation and electric energy distribution problems can be caused. Aiming at the problem of power fluctuation and considering economic benefits, a household energy management method considering household photovoltaic power generation is established in a household daily load energy model by combining a multi-step graph neural network self-adaptive dynamic planning method. Firstly, a load electrical model is established based on actual data, and the load electrical model can reflect the load electricity utilization characteristics of residents. And then, building a resident load behavior model, and obtaining a family load size probability model for each hour based on Weibull distribution. And finally, obtaining a daily load curve of the single family through the synthesis of the two. The load electrical model taken here is an exponential model:
Figure BDA0002659667750000021
in the formula: psActive power for the load; u is a supply voltage; p0Rating the active power for the load; u shape0Representing the nominal voltage magnitude of the system. Z is a linear or branched memberp,Ip,PpFor the constant parameter to be solved, the power exponent α can be calculated by the equation (2):
Figure BDA0002659667750000022
in the load behavior model, the number n of effective consumers of electricity must be determined firsttAs shown in formula (3):
Figure BDA0002659667750000023
in the formula: n is the number of family members; w is 0 for weekend, w is 1 for workday; pi,jIs the probability from state i to state j. The state transition probability is calculated as shown in (4):
Figure BDA0002659667750000024
in the formula: n isi,jIs the number of events that transition from i to j; n represents the total number of states.
And the switch state S of the resident load1Determined as equation (5):
Figure BDA0002659667750000025
in the formula: n (t) is the number of effective electricity utilization people of residents at the moment t; p is1(t) is the daily load utilization probability of residents; 1 is on and 0 is off. And finally, obtaining a family daily load probability model through a Weibull probability density function.
A photovoltaic power generation management method of multi-step graph neural network adaptive dynamic planning comprises three basic modules, namely a graph model network module, a graph evaluation network module and a graph control network module, wherein the core structure of each basic module is a graph neural network. The graph neural network forms a visual picture of a large amount of photovoltaic power generation data and establishes a photovoltaic power generation data model, and the implementation steps are as follows:
(1) the graph neural network firstly extracts data characteristics from photovoltaic power generation data:
hν＝f(xν,xco[ν],hne[ν],xne[ν]) (6)
and oν＝g(hν,xν) V denotes a node; h isνRepresents state embedding, contains domain information for each node, is an s-dimensional vector of node v, which can be used to generate output omicronν(ii) a f is a parameter function, is shared among all nodes, and updates the node state according to the input field; g is another parametric function describing how the output is generated;
xν，xco[ν]，hne[ν]and xne[ν]Respectively is the characteristics of the v node, the characteristics of the edge, the state and the characteristics of the v adjacent node;
(2) initializing the representation of the nodes by using the data characteristics, and establishing a photovoltaic power generation data model of the graph neural network:
Figure BDA0002659667750000031
Figure BDA0002659667750000032
Figure BDA0002659667750000033
Figure BDA0002659667750000034
Figure BDA0002659667750000035
Aνrepresenting the connection of node v to its neighbors, hνRepresenting the state of node v, b represents the bias;
Figure BDA0002659667750000036
collecting neighborhood information of a node v, wherein z and r are an updating gate and a resetting gate respectively; sigma and tanh respectively represent a Sigmoid function and a hyperbolic tangent function, and represent element multiplication operation; w, U, WZ，UZ，WrAnd UrIs a parameter matrix to be learned;
(3) and (3) propagating data information in the graph, and obtaining a photovoltaic power generation data model through a complete connection layer:
Figure BDA0002659667750000037
in order to enable the power fluctuation to be as small as possible and the economic benefit to be as large as possible, the multi-step self-adaptive dynamic programming method regulates and controls the charging and discharging capacity of the energy storage unit, reduces the household load fluctuation as much as possible, and simultaneously ensures the best economic benefit. The objective function of the multi-step self-adaptive dynamic programming method is set as follows:
Figure BDA0002659667750000038
Figure BDA0002659667750000041
in formulae (6) and (7):
Figure BDA0002659667750000042
is the average value of the daily load of the family; pH(t) is a small value of the adjusted daily load of the family; pp(t) energy storage per hour; and S (t) is the corresponding real-time electricity price which is divided into a peak value and a valley value.
The household energy cannot exceed the coverage of the battery capacity. Therefore, the constraint is:
0≤P(t)-PH(t)≤Ppmax (14)
in the formula: p (t) is the real-time power value before adjustment.
The photovoltaic power generation data model and the family daily load probability model are used as the input of the multi-step self-adaptive dynamic planning, and the method comprises the following specific steps:
the first step is as follows:
acquiring the current state delta f1 of the photovoltaic power generation system;
inputting x (k) into the graph control network module 1, and outputting a control action u (k);
inputting x (k) and u (k) into a graph evaluation network module 0, and outputting J [ x (k) ];
inputting x (k) and u (k) into a graph model network module 1, outputting x (k +1), and predicting the state at the next moment;
inputting x (k +1) into a graph evaluation network module 1, and outputting J [ x (k +1) ];
the second step is that:
acquiring a state delta f2 of a photovoltaic power generation system x (k + 1);
inputting x (k +1) into a graph control network module 2, and outputting a control action u (k + 1);
inputting x (k +1) and u (k +1) into a graph model network module 2, outputting x (k +2), and predicting the state at the next moment;
inputting x (k +2) into a graph evaluation network module 2, and outputting J [ x (k +2) ];
the nth step:
acquiring a state delta fn when a photovoltaic power generation system x (k + n-1) is in a state;
inputting x (k + n-1) into a graph execution network module n, and outputting a control action u (k + 1);
inputting x (k + n-1) and u (k + n-1) into a graph model network module n, outputting x (k + n), and predicting the state at the next moment;
inputting x (k + n) into a graph evaluation network module n, and outputting J [ x (k + n) ];
after the multi-step self-adaptive dynamic planning method is carried out, the control strategy of the photovoltaic power generation system can be updated on line according to the output, and the state of the system at the next moment is predicted. And carrying out multi-objective planning according to daily load fluctuation and electricity purchasing and selling cost, storing energy for photovoltaic power generation which is not consumed in the daytime through the energy storage unit, releasing electric energy at night for power supply, and carrying out grid-connected selling on the residual electricity. And optimally controlling the multi-target planning problem of photovoltaic power generation energy management by a multi-step graph neural network adaptive dynamic planning method.
Drawings
FIG. 1 is a schematic structural diagram of a multi-step diagram neural network adaptive dynamic planning method of the present invention.
FIG. 2 is a flow chart of the photovoltaic power generation management of the multi-step graph neural network adaptive dynamic programming of the method of the present invention.
Detailed Description
The invention provides a photovoltaic power generation management method adopting a multi-step graph neural network self-adaptive dynamic planning, which is described in detail in combination with the attached drawings as follows:
FIG. 1 is a schematic structural diagram of a multi-step diagram neural network adaptive dynamic planning method of the present invention. The method mainly comprises three basic modules, namely a graph control network module, a graph model network module and a graph evaluation network module, wherein the core structure of each module is a graph neural network. The multi-step graph neural network self-adaptive dynamic planning method is based on the neural network self-adaptive dynamic planning, and the graph neural network is integrated into the neural network dynamic planning method to form the graph neural network self-adaptive dynamic planning method. The traditional neural network self-adaptive dynamic planning method adopts a neural network mode to fit or approximate a Hamilton-Jacobian-Bellman equation, all basic artificial neural networks are realized, and when the complexity of a system becomes high, a system model cannot be obtained or the system has a serious hysteresis loop, the traditional artificial neural networks are difficult to approximate a system function, so that a more intelligent self-adaptive dynamic planning method is necessary to be designed. Due to the addition of a multi-step strategy, the prediction capability of the system is further improved, and the convergence process is accelerated; and the learning result can be updated in real time, which is beneficial to the improvement of system control and optimization performance.
FIG. 2 is a flow chart of the photovoltaic power generation management of the multi-step graph neural network adaptive dynamic programming of the method of the present invention. The multi-step graph neural network self-adaptive dynamic planning method comprises a graph neural network method and a multi-step self-adaptive dynamic planning method, wherein the graph neural network acts on a large amount of photovoltaic power generation data to form a visual picture, and a photovoltaic power generation data model is established:
Figure BDA0002659667750000051
in the formula: h isνRepresents state embedding, contains domain information for each node, is an s-dimensional vector of node v, which can be used to generate output omicronν；xνRepresenting the characteristics of node v.
Combining a load electrical model and a resident load behavioral model based on Weibull distribution to obtain a family daily load probability model, wherein the load electrical model is an exponential model:
Figure BDA0002659667750000052
in the formula: psActive power for the load; u is a supply voltage; p0Rating the active power for the load; u shape0Representing the nominal voltage magnitude of the system. Zp,IpAnd PpFor a constant parameter to be solved, the power exponent α can be calculated as:
Figure BDA0002659667750000061
in the load behavior model, the number n of effective consumers of electricity must be determined firsttAs shown in the following formula:
Figure BDA0002659667750000064
in the formula: n is the number of family members; w is 0 for weekend, w is 1 for workday; pi,jIs the probability from state i to state j. The state transition probabilities are calculated as follows:
Figure BDA0002659667750000062
in the formula: n isi,jIs the number of events that transition from i to j; n represents the total number of states.
And the switch state S of the resident load1It can be determined as:
Figure BDA0002659667750000063
in the formula: n (t) is the number of effective electricity utilization people of residents at the moment t; p1(t) is the daily load utilization probability of residents; 1 is on and 0 is off.
The photovoltaic power generation data model and the family daily load probability model are used as input of the multi-step self-adaptive dynamic planning, the control strategy of the photovoltaic power generation system can be updated on line according to output, and the state of the system at the next moment is predicted. And carrying out multi-objective planning according to daily load fluctuation and electricity purchasing and selling cost, storing energy for photovoltaic power generation which is not consumed in the daytime through the energy storage unit, releasing electric energy at night for power supply, and carrying out grid-connected selling on the residual electricity. And optimally controlling the multi-target planning problem of photovoltaic power generation energy management by a multi-step graph neural network adaptive dynamic planning method.

Claims (1)

1. A photovoltaic power generation management method of a multi-step graph neural network self-adaptive dynamic planning is characterized in that a graph neural network is added on the basis of the self-adaptive dynamic planning based on the neural network, a large amount of photovoltaic power generation data can be processed visually in a picture mode, a multi-step self-adaptive dynamic planning method is added, the photovoltaic power generation data are controlled and optimized in real time, and the system is predicted at the next moment; the method comprises the following steps in the using process:
(1) the graph neural network method is acted on a large amount of photovoltaic power generation data to form a visual picture, and a photovoltaic power generation data model is established; the graph neural network forms a visual picture of a large amount of photovoltaic power generation data, a photovoltaic power generation data model is established, and the working process based on the graph neural network is as follows:
1) the graph neural network firstly extracts data characteristics from photovoltaic power generation data:
hν＝f(xν,xco[ν],hne[ν],xne[ν])
oν＝g(hν,xν)
v represents a node; h isνRepresenting state embedding, including domain information for each node, is an s-dimensional vector of node v, used to generate output oν(ii) a f is a parameter function, is shared among all nodes, and updates the node state according to the input field; g is another parametric function; x is the number ofν、xco[ν]、hne[ν]And xne[ν]Respectively are the characteristics of the v node, the characteristics and the state of the edge and the characteristics of the v adjacent node;
2) initializing the representation of the nodes by using the data characteristics, and establishing a photovoltaic power generation data model of the graph neural network:
Figure FDA0003584865330000011
Figure FDA0003584865330000012
Figure FDA0003584865330000013
Figure FDA0003584865330000014
Figure FDA0003584865330000015
Aνrepresenting the connection of the node v with its neighboring nodes; h is a total ofνRepresenting the state of a node v; b represents a bias;
Figure FDA0003584865330000016
collecting neighborhood information of the node v; z and r are the update gate and the reset gate, respectively; sigma and tanh respectively represent a Sigmoid function and a hyperbolic tangent function, and represent element multiplication operation; w, U, WZ，UZ，WrAnd UrIs a parameter matrix to be learned;
3) and (3) propagating data information in the graph, and obtaining a photovoltaic power generation data model through a complete connection layer:
Figure FDA0003584865330000017
obtaining the final representation of the node and forming a visual picture;
(2) carrying out real-time analysis and optimization on the photovoltaic power generation system and predicting the next time of the system by using a multi-step self-adaptive dynamic programming method, and outputting a control signal; the multi-step self-adaptive dynamic planning method consists of three basic modules, namely a graph model network module, a graph evaluation network module and a graph control network module, and comprises the following steps:
the first step is as follows:
acquiring the current state delta f1 of the photovoltaic power generation system;
inputting x (k) into the graph control network module 1, and outputting a control action u (k);
inputting x (k) and u (k) into a graph evaluation network module 0, and outputting J [ x (k) ];
inputting x (k) and u (k) into a graph model network module 1, outputting x (k +1), and predicting the state at the next moment;
inputting x (k +1) into a graph evaluation network module 1, and outputting J [ x (k +1) ];
the second step is that:
acquiring a state delta f2 of a photovoltaic power generation system x (k + 1);
inputting x (k +1) into a graph control network module 2, and outputting a control action u (k + 1);
inputting x (k +1) and u (k +1) into a graph model network module 2, outputting x (k +2), and predicting the state at the next moment;
inputting x (k +2) into a graph evaluation network module 2, and outputting J [ x (k +2) ];
the nth step:
acquiring a state delta fn when a photovoltaic power generation system x (k + n-1) is in a state;
inputting x (k + n-1) into a graph execution network module n, and outputting a control action u (k + 1);
inputting x (k + n-1) and u (k + n-1) into a graph model network module n, outputting x (k + n), and predicting the state at the next moment;
inputting x (k + n) into a graph evaluation network module n, and outputting J [ x (k + n) ];
after the multi-step self-adaptive dynamic planning algorithm is carried out, the control strategy of the photovoltaic power generation system can be updated on line, the state of the system at the next moment is predicted, and the multi-target planning problem of photovoltaic power generation energy management is optimally controlled.
CN202010900514.8A 2020-09-01 2020-09-01 Photovoltaic power generation management method for multi-step graph neural network self-adaptive dynamic planning Active CN112217198B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010900514.8A CN112217198B (en) 2020-09-01 2020-09-01 Photovoltaic power generation management method for multi-step graph neural network self-adaptive dynamic planning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010900514.8A CN112217198B (en) 2020-09-01 2020-09-01 Photovoltaic power generation management method for multi-step graph neural network self-adaptive dynamic planning

Publications (2)

Publication Number Publication Date
CN112217198A CN112217198A (en) 2021-01-12
CN112217198B true CN112217198B (en) 2022-05-17

Family

ID=74048720

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010900514.8A Active CN112217198B (en) 2020-09-01 2020-09-01 Photovoltaic power generation management method for multi-step graph neural network self-adaptive dynamic planning

Country Status (1)

Country Link
CN (1) CN112217198B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116526567A (en) * 2023-06-26 2023-08-01 深圳智电时代科技有限公司 Energy management method and device for household photovoltaic power supply system and storage medium

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110406526A (en) * 2019-08-05 2019-11-05 合肥工业大学 Parallel hybrid electric energy management method based on adaptive Dynamic Programming

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1317673C (en) * 2004-03-18 2007-05-23 致伸科技股份有限公司 System and method for distinguishing words and graphics in an image using neural network
US10789530B2 (en) * 2019-01-14 2020-09-29 Capital One Services, Llc Neural embeddings of transaction data
CN110188637A (en) * 2019-05-17 2019-08-30 西安电子科技大学 A kind of Activity recognition technical method based on deep learning
CN111275256A (en) * 2020-01-18 2020-06-12 杭州电子科技大学 Photovoltaic power generation power day-ahead prediction method based on image feature extraction

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110406526A (en) * 2019-08-05 2019-11-05 合肥工业大学 Parallel hybrid electric energy management method based on adaptive Dynamic Programming

Also Published As

Publication number Publication date
CN112217198A (en) 2021-01-12

Similar Documents

Publication Publication Date Title
CN110119886B (en) Active distribution network dynamic planning method
Berrazouane et al. Parameter optimization via cuckoo optimization algorithm of fuzzy controller for energy management of a hybrid power system
CN107634518B (en) Source-network-load coordinated active power distribution network economic dispatching method
CN107545325A (en) A kind of more microgrid interconnected operation optimization methods based on game theory
CN112713618B (en) Active power distribution network source network load storage cooperative optimization operation method based on multi-scene technology
CN107994595A (en) A kind of system of peak load shifting control method and system and the application control method
CN109599856A (en) Electric car management of charging and discharging optimization method and device in a kind of more building of microgrid
Chen et al. Multi-time scale coordinated optimal dispatch of microgrid cluster based on MAS
Zhu et al. A parallel meta-heuristic method for solving large scale unit commitment considering the integration of new energy sectors
Vosoogh et al. An intelligent day ahead energy management framework for networked microgrids considering high penetration of electric vehicles
Xu et al. Energy allocation strategy based on fuzzy control considering optimal decision boundaries of standalone hybrid energy systems
CN113224769B (en) Multi-time-scale power distribution network voltage optimization method considering photovoltaic multi-state adjustment
CN111585279B (en) Micro-grid optimal scheduling method based on new energy consumption
CN109901389A (en) A kind of new energy consumption method based on deep learning
CN116667325A (en) Micro-grid-connected operation optimization scheduling method based on improved cuckoo algorithm
CN112217198B (en) Photovoltaic power generation management method for multi-step graph neural network self-adaptive dynamic planning
Kowahl et al. Micro-scale smart grid optimization
CN113723793A (en) Method, device, equipment and medium for realizing park comprehensive energy system
CN113469839A (en) Smart park optimization strategy based on deep reinforcement learning
CN116865270A (en) Optimal scheduling method and system for flexible interconnection power distribution network containing embedded direct current
CN106600078A (en) Micro-grid energy management scheme based on new energy power generation forecasting
Li et al. Adaptive dynamic programming approach for micro-grid optimal energy transmission scheduling
CN115411776A (en) Thermoelectric cooperative scheduling method and device for residential comprehensive energy system
CN108512237A (en) Light based on intelligent fuzzy decision stores up association system real-time scheduling method
KR102562761B1 (en) Method for intelligent day-ahead energy sharing scheduling of the P2P prosumer community in smart grid

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant